Missions Overview

Missions are the core unit of work in Buildd. Each mission represents a task that an AI agent will claim, plan, and execute. Missions come in three types, each designed for a different workflow.

Build Missions

Build missions are the most common type. They produce code changes — features, bug fixes, refactors, or any work that results in a pull request.

When a worker claims a Build mission, it:

1. Analyzes the task description and context
2. Creates a plan (optionally reviewed via Planning Mode)
3. Makes code changes in the target repository
4. Opens a pull request with the results

Build missions are best for: new features, bug fixes, refactors, dependency upgrades, and test additions.

Watch Missions

Watch missions run on a schedule and monitor your codebase or infrastructure for issues. They don't produce code changes by default — instead, they report observations and can escalate to Build missions when action is needed.

Watch missions are best for: monitoring CI failures, tracking dependency vulnerabilities, reviewing error logs, and enforcing code quality standards.

See Schedules for configuring recurring Watch missions.

Brief Missions

Brief missions produce non-code deliverables — analysis documents, architecture reviews, research reports, or any output that's better expressed as a written artifact rather than a pull request.

When a worker completes a Brief mission, the output is saved as an Artifact with a shareable link.

Brief missions are best for: architecture reviews, competitive analysis, documentation drafts, and technical research.

Creating Missions

You can create missions through:

• Dashboard — The web UI at buildd.dev
• MCP Server — Directly from Claude Code via the MCP integration
• API — Programmatically via POST /api/tasks
• GitHub — Automatically from issues via GitHub Integration

Mission Lifecycle

Every mission follows the same lifecycle:

1. Created — Mission is defined with a description, type, and optional context
2. Queued — Mission is waiting for an available worker
3. Claimed — A worker has picked up the mission
4. In Progress — The worker is actively executing
5. Waiting Input — The worker needs human input (e.g., plan approval)
6. Completed — The mission is done, with results available
7. Failed — The mission encountered an unrecoverable error

Workers can attach Memory entries during execution, building a persistent knowledge base that improves future missions.